It seems that even in the cutting-edge world of AI development, good old-fashioned user frustration can still make waves. Recently, developers using Google's Antigravity AI programming tool have found themselves in a bit of a bind, voicing their complaints about unexpected price hikes and changes to their access. It’s a situation that feels all too familiar, doesn't it? When something we rely on suddenly shifts, especially when it impacts our workflow and budget, it’s natural to feel a bit… unsettled.
Google announced on X (formerly Twitter) that they were "optimizing their AI plans to give users more control over what they can build." A noble goal, certainly. The update introduced a new system where AI credits are now the currency for Antigravity. Subscriptions come with some built-in credits, and you can buy more – 2500 credits for $25, to be exact. The catch? The exact value of a single credit within Antigravity wasn't immediately clear, leaving many users scratching their heads.
The developer forums, particularly for those on the $20/month AI Pro subscription, quickly filled with grumbles. This plan, as described at the time, promised "high, generous quotas that refresh every five hours until you hit your weekly limit." But users reported that this generous refresh rate seemed to have vanished, replaced by a weekly refresh instead. For developers who depend on continuous access, this meant hitting a wall and being unable to continue their work unless they bought more credits or upgraded to a pricier plan.
Antigravity itself supports a range of powerful large language models, including different configurations of Gemini 3.1 Pro, Gemini 3 Flash, Anthropic's Claude Sonnet 4.6 and Claude Opus 4.6, and even OpenAI's GPT-OSS 120B. Flash is positioned as the efficient, budget-friendly option, while the Pro plan is aimed at "enthusiasts, students, and developers working in an IDE." For the heavy hitters, the $249.99/month AI Ultra plan offers "continuous, high-volume access to our most complex models."
One developer on Reddit shared a stark comparison: "Before January, I could use the Gemini Pro model with over 300 million input/1-2 million output tokens weekly," they stated. "This week, I hit the weekly limit with less than 9 million input/200,000 output tokens." That's a significant drop, and it’s easy to see why people are upset. Antigravity first launched in preview in November 2025 without pricing, and early descriptions of quotas used vague terms like "high" and "generous," making it difficult for users to truly understand their limits.
Adding another layer to the user discontent, a separate issue emerged where Google temporarily blocked users from Antigravity who were accessing Google's Gemini models via a third-party tool called OpenClaw. Google cited "large-scale malicious use" on the Antigravity backend, which was allegedly causing significant strain and impacting service quality for legitimate users. OpenClaw, it turns out, was using Antigravity's backend as a proxy, leading to unexpectedly high computational loads. While Google stated they would offer a path to restore access for some users who might not have realized they were violating terms of service, the sudden bans, especially for paid users, caused considerable alarm and a feeling of being unfairly treated.
This situation highlights a broader tension in the AI industry: the balance between fostering an open ecosystem and maintaining control over services, pricing, and user experience. As AI tools become more integrated into professional workflows, transparency and predictable access become paramount. When these expectations are unmet, the result is often a chorus of user complaints, echoing the sentiment that while innovation is exciting, it shouldn't come at the cost of clear communication and fair treatment.
